"It's not me, but rather the circumstances we find ourselves in, that "insist" on boxical evidence. As you say later, it is "the system we are locked into"."

But what you seem to be doing is begging the question and arguing in a circle.  It is all well and good to observe that we find ourselves locked inside a box, but as soon as you ask the question "but is the box the sum of all reality?" you cannot now answer, "yes, because we know the box is the sum of all reality."  In order to escape the charge of circular reasoning and begging the question, the answer has to have a different form:  "yes, because such and such"

We may observe that we are precluded from exiting our box in order to make direct inquiries on our own, and this observation may be permitted a certain amount of weight towards the answer, "yes it is."  However, if we are reasonable chaps, having determined that direct observation isn't viable, we will now consider whether or not there might be other ways to tackle the question.

We may reasonably hypothesize that one such way that we could learn about superboxical realities is for a superboxical reality to come into our box and make known certain facts.  In doing so, it would be employing revelation.

But we have observed that even in that situation, the revelation is coming to us through boxical means of sound waves, light, etc, things that we consider boxical.   Given this, we must allow two things:  1., if the superboxical entity is going to substantiate his claim, its going to have to be carefully crafted to actually do so and 2., we ourselves will have to prepare ourselves with a list of things that we could understand would reasonably substantiate his claim.

It is precisely on these two questions that I am dwelling on.

"Yes, all evidence is physical evidence, but the laws of physics are immutable, so any demonstrable suspension of said laws is strong evidence in favour of a non-physical phenomenon (or entity or force or whatever)."

I'm glad you said that.  It's in the category of #2 above, but note I said 'reasonably substantiate.'  :)  Basically you are asserting that a miracle would be evidence for God.   No objection to that.  But where did you get your knowledge that the 'laws of physics are immutable'?  How did you determine that?  And how did you determine the 'laws of physics' in the first place?

"I'm happy to eliminate the possibility of deceptive testimony for the sake of argument"

I understand that this is what you are trying to do, but I think it is very important that we understand precisely what testimony really is.  As you explain, someone takes their memories, processes them, and shares them with someone else.  On your view, that is an entirely physical process, no?  dancing.... dancing...

"Now suppose that someone comes along and says "I saw a pink elephant!"."

My problem with your analogy is that you do not specify that there are, in fact, no such thing as pink elephants.  :)  Your example takes it for granted that they don't exist and that I also understand that they don't exist.  If its just a matter of reliability, and the pitfalls associated with testimony, I don't have to be persuaded.  I really don't.  However, the example you choose to give hints at how you perceive the question of God.  Just as we 'know' there are no pink elephants and have to find alternative explanations for why truthful people will still report them, you imply that we 'know' there is no God, so, we have to explain why people still testify to him- but of course, with them the problem if deception is back in the mix.    We know there is no God, so we cannot trust those who testify about him.  That seems to be your argument.

A much better example that relates to what we're debating would have been if someone came and told you that they saw a hutriacartic marpholopolous.

How would you go about validating THAT claim?

"So yes, you can say that the testimony is evidence of a memory, and the memory is evidence of some past event, but, in assessing the worth of that evidence, we must consider the reliability of the mechanisms used to encode, store and retrieve it, not to mention the existence of other corroborating evidence that is not memory-based."

Dancing.... dancing... Tim's dancing on top of Sntjohnny's Golden Rule of Epistemology.... dancing... dancing... :)

 

Food for thought...
http://www.smh.com.au/world/science/memories-you-cannot-swear-by-20110914-1k9k5.html

Something I had already heard about, and interesting- although I just said above that I didn't need to be persuaded of such things- but in context of this discussion I feel compelled now to challenge the validity of the studies. 

Example:

A week later they were asked what happened. They were also asked whether they "remembered" or simply "knew" facts, to see if they differentiated between what they had seen and what they had been told.

The results were worrying.

How could they be 'worrying'?  This presumes that the persons carrying out the studies remembered correctly what happened the previous week.  It even presumes that the persons carrying out the studies remembered which study they were involved in.  We may suppose perhaps that one of the study authors was actually an auto mechanic that forgot his way to the shop and wandered into the classroom, where the other study authors couldn't remember whether or not the mechanic was present the week before.  Probably because they weren't, either!  The actual study authors from the week before forgot the project that they were involved in.  Fortunately, one of them ended up at the auto shop and people were still able to get their oil changed!